Online monitoring method and device for circuit breaker in power grid and electronic equipment

Abstract

The invention discloses an online monitoring method and device for a circuit breaker in a power grid and electronic equipment. The method comprises the steps that in the operation process of a target circuit breaker in a power grid, sound data and vibration data generated by the target circuit breaker are acquired, and the target circuit breaker is used for representing the circuit breaker needing online monitoring in the power grid; a target circuit breaker state detection model is used for monitoring the operation state of the target circuit breaker based on the sound data and the vibration data, a monitoring result is obtained, and the target circuit breaker state detection model is obtained through pre-training based on sample sound data and sample vibration data of the circuit breaker in the normal operation state; and the machine learning model is used for monitoring the state of the circuit breaker, and the monitoring result is used for representing whether the target circuit breaker is in a fault state. According to the invention, the technical problem of low efficiency of condition detection of the circuit breaker in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of circuit breaker monitoring, and in particular to an online monitoring method, device and electronic equipment for a circuit breaker in a power grid. Background Art

[0002] Monitor the circuit breakers in the power grid, that is, check and evaluate the working status and performance of the circuit breakers to ensure the normal operation of the circuit breakers and protect the safe and stable operation of the power grid. By monitoring the operating status of the circuit breakers, abnormal conditions of the circuit breakers can be discovered in time, including overload, short circuit, poor contact and other problems, so that timely measures can be taken to repair them and ensure the normal operation of the power grid.

[0003] At present, the circuit breaker monitoring methods in related technologies are mostly used in offline live detection or power outage detection scenarios, or when performing condition detection and fault diagnosis on circuit breakers, it is necessary to obtain relevant data of circuit breakers in a faulty state in advance as a data basis. Data acquisition is relatively inconvenient, resulting in low efficiency in condition detection of circuit breakers in related technologies.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method, a device and an electronic device for online monitoring of a circuit breaker in a power grid, so as to at least solve the technical problem of low efficiency in condition detection of a circuit breaker in the related art.

[0006] According to one aspect of an embodiment of the present invention, there is provided an online monitoring method for a circuit breaker in a power grid, comprising: acquiring sound data and vibration data generated by a target circuit breaker during operation of the target circuit breaker in the power grid, wherein the target circuit breaker is used to represent a circuit breaker in the power grid that needs to be monitored online; and monitoring the operating state of the target circuit breaker based on the sound data and the vibration data using a target circuit breaker state detection model to obtain a monitoring result, wherein the target circuit breaker state detection model is used to represent a machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of the circuit breaker in a normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state.

[0007] Optionally, the target circuit breaker state detection model is used to monitor the operating state of the target circuit breaker based on the sound data and the vibration data to obtain a monitoring result, including: extracting features from the sound data to obtain a sound voiceprint feature, and extracting features from the vibration data to obtain a vibration voiceprint feature; fusing the sound voiceprint feature and the vibration voiceprint feature to obtain a fused voiceprint feature; and using the target circuit breaker state detection model to monitor the operating state of the target circuit breaker based on the fused voiceprint feature to obtain a monitoring result.

[0008] Optionally, the method also includes: when the monitoring result shows that the target circuit breaker is not in a fault state, continuously performing online monitoring on the target circuit breaker; when the monitoring result shows that the target circuit breaker is in a fault state, obtaining the coil current signal and the moving contact travel signal of the target circuit breaker; and determining the fault type of the target circuit breaker based on the coil current signal and the moving contact travel signal.

[0009] Optionally, the method also includes: obtaining sample sound data and sample vibration data of at least one circuit breaker in normal operating state; constructing an initial voiceprint feature vector of a target circuit breaker state detection model based on the sample sound data and the sample vibration data; performing dimensionality reduction processing on the initial voiceprint feature vector to obtain a target voiceprint feature vector, wherein the intra-class distance of the target voiceprint feature vector is smaller than the intra-class distance of the initial voiceprint feature vector, and the inter-class distance of the target voiceprint feature vector is larger than the inter-class distance of the initial voiceprint feature vector, and constructing a target circuit breaker state detection model based on the target voiceprint feature vector.

[0010] Optionally, based on the sample sound data and the sample vibration data, constructing an initial voiceprint feature vector of the target circuit breaker state detection model, including: performing short-time Fourier transform on the sample sound data and the sample vibration data to obtain multiple time-frequency spectrum feature vectors; constructing a training sample set based on the multiple time-frequency spectrum feature vectors; based on the training sample set and a preset deep convolutional neural network, constructing an initial voiceprint feature vector, wherein the preset deep convolutional neural network is used to represent a preset basic deep convolutional neural network.

[0011] Optionally, constructing an initial voiceprint feature vector based on a training sample set and a preset deep convolutional neural network includes: determining a data scale of the training sample set, wherein the data scale includes at least one of the following: a frame length and a frame shift of multiple time-frequency spectrum feature vectors and the number of frames for constructing multiple time-frequency spectrum feature vectors of the training sample set; determining model parameters for constructing a target circuit breaker state detection model based on the data scale, wherein the model parameters include at least one of the following: the number of layers of the target circuit breaker state detection model, the size of convolution kernels of each layer and the number of convolution kernels of each layer; constructing an initial voiceprint feature vector based on the model parameters and a preset deep convolutional neural network.

[0012] According to another aspect of an embodiment of the present invention, there is further provided an online monitoring device for a circuit breaker in a power grid, comprising: an acquisition module, for acquiring sound data and vibration data generated by a target circuit breaker during operation of the target circuit breaker in the power grid, wherein the target circuit breaker is used to represent a circuit breaker that needs to be monitored online in the power grid; a monitoring module, for monitoring the operating state of the target circuit breaker based on the sound data and the vibration data using a target circuit breaker state detection model to obtain a monitoring result, wherein the target circuit breaker state detection model is used to represent a machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of the circuit breaker in a normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state.

[0013] In an embodiment of the present invention, an online monitoring method for a circuit breaker in a power grid is provided, comprising: during the operation of a target circuit breaker in the power grid, obtaining sound data and vibration data generated by the target circuit breaker, wherein the target circuit breaker is used to represent a circuit breaker that needs to be monitored online in the power grid; using a target circuit breaker state detection model to monitor the operating state of the target circuit breaker based on the sound data and the vibration data to obtain a monitoring result, wherein the target circuit breaker state detection model is used to represent a machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of the circuit breaker in a normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state. It is easy to notice that the present application can realize real-time monitoring of the operating status of the target circuit breaker by collecting the sound data and vibration data generated by the target circuit breaker and monitoring it in combination with the target circuit breaker status detection model. The monitoring method based on sound data and vibration data can more intuitively reflect the operating status of the target circuit breaker and realize online monitoring of the target circuit breaker without offline or power-off inspection of the circuit breaker. At the same time, it can also improve the accuracy and efficiency of monitoring, and can timely discover whether the target circuit breaker is in a fault state, so as to take maintenance or repair measures in time, thereby solving the technical problem of low efficiency of condition detection of circuit breakers in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0015] Figure 1 is a flow chart of an online monitoring method for a circuit breaker in a power grid according to an embodiment of the present invention;

[0016] Figure 2 is a flowchart of an optional method for generating a target voiceprint feature vector according to an embodiment of the present invention;

[0017] Figure 3 is a schematic diagram of an optional online monitoring process of a circuit breaker in a power grid according to an embodiment of the present invention;

[0018] Figure 4 Schematic diagram of an online monitoring device for a circuit breaker in a power grid according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0020] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] According to an embodiment of the present invention, an embodiment of an online monitoring method for a circuit breaker in a power grid is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0022] Figure 1 is a flow chart of an online monitoring method for a circuit breaker in a power grid according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0023] Step S102: Acquire sound data and vibration data generated by the target circuit breaker during operation of the target circuit breaker in the power grid.

[0024] The target circuit breaker is used to represent the circuit breaker that needs to be monitored online in the power grid.

[0025] The above-mentioned target circuit breaker may refer to a device in a power grid used to protect a circuit from overload and short circuit, which can automatically cut off the circuit when the current exceeds a set value to prevent circuit damage or fire. The target circuit breaker may be one or more circuit breakers in the power grid. The specific type of the target circuit breaker may be determined according to actual needs and is not limited here.

[0026] The above-mentioned sound data may refer to the sound signal generated by the circuit breaker during operation. The sound signal may include various frequencies, amplitudes and durations. By analyzing the sound data, it is possible to determine whether the circuit breaker is operating normally, whether there is a fault and the specific type of the fault. When the circuit breaker is operating normally, a stable and regular sound signal will be generated, and the frequency and amplitude will also change within a relatively stable range. When the circuit breaker fails, such as partial discharge, poor contact and other problems, the frequency and amplitude of the sound signal will change abnormally. The operating status of the circuit breaker can be determined by analyzing these abnormally changing sound data. The specific type of sound data can be determined according to actual needs and is not limited here.

[0027] The above-mentioned vibration data may refer to the vibration signal data generated by the circuit breaker during operation. The vibration data may reflect the internal operating status and performance of the circuit breaker, including the opening and closing process of the circuit breaker, the movement status of mechanical parts, the degree of bearing wear, looseness, etc. For example, when the circuit breaker is opened and closed, a series of vibration signals may be generated. By monitoring and analyzing the vibration signals, it is possible to judge the opening and closing speed of the circuit breaker, whether the movement is smooth, whether the mechanical parts are operating normally, etc. If there is a fault or abnormality inside the circuit breaker, it will cause changes in the vibration signal. By monitoring these changes, the problem can be discovered and located in time, and corresponding maintenance and repair work can be performed.

[0028] In an optional embodiment, the sound data and vibration data generated by the target circuit breaker can be obtained by a sensor device, and the sensor device can be a sound sensor and a vibration sensor. The sound and vibration signals of the target circuit breaker can be collected in real time by the sensor device. Specifically, according to the characteristics of the target circuit breaker and the monitoring requirements, a suitable sound sensor and a vibration sensor can be selected, and the sensor device needs to be able to accurately and stably collect the sound and vibration signals generated by the target circuit breaker; the selected sound sensor and vibration sensor can be installed near the target circuit breaker or directly fixed on the target circuit breaker to ensure that the distance between the sensor and the target circuit breaker is appropriate; the sensor device can be connected to a data acquisition system to ensure that the sensor device can transmit the collected sound and vibration data to the data acquisition system; the sound and vibration data collected by the sensor can be monitored and recorded in real time by the data acquisition system; the sound data and vibration data can also be directly retrieved from the database system of the power grid. The method of obtaining the sound data and vibration data can be determined according to actual needs and is not limited here. The obtained sound data and vibration data can be used for subsequent online monitoring of the target circuit breaker.

[0029] Step S104: monitor the operating state of the target circuit breaker based on the sound data and the vibration data using the target circuit breaker state detection model to obtain a monitoring result.

[0030] Among them, the target circuit breaker state detection model is used to represent a machine learning model for circuit breaker state monitoring that is pre-trained based on sample sound data and sample vibration data of a circuit breaker in normal operating status. The monitoring result is used to indicate whether the target circuit breaker is in a fault state.

[0031] The above-mentioned target circuit breaker state detection model may refer to a machine learning model pre-trained based on sample sound data and sample vibration data of a circuit breaker in normal operating state, which can be used to monitor the operating state of the target circuit breaker, and determine whether the target circuit breaker is in a fault state and determine the type of fault based on the sound and vibration data.

[0032] The present application can obtain the sound and vibration data of the circuit breaker in normal operation, and use these data to train the target circuit breaker state detection model. Through training, the model can learn the sound and vibration characteristics of the circuit breaker in normal state. When the target circuit breaker is in operation, the monitoring system can collect the sound and vibration data of the target circuit breaker in real time, and input it into the target circuit breaker state detection model for analysis. The model will extract and compare the features of these data to determine whether the operating state of the target circuit breaker is normal. If the model detects an abnormality, it will output the corresponding monitoring results, indicating that the target circuit breaker may be in a fault state and requires further inspection or maintenance.

[0033] The above-mentioned monitoring results may refer to the results obtained after monitoring the operating status of the target circuit breaker based on sound data and vibration data through the target circuit breaker status detection model, which may include whether the target circuit breaker is in a normal operating state or a fault state, and the specific type of fault, such as overload, short circuit, etc. It may also include a health status assessment of the target circuit breaker, such as predicting the life of the circuit breaker or recommending maintenance measures.

[0034] In an optional embodiment, the target circuit breaker state detection model can be used to monitor the operating state of the target circuit breaker based on sound data and vibration data. The collected sound data and vibration data need to be preprocessed, including denoising, filtering, downsampling and other operations to ensure the quality and accuracy of the data; the trained target circuit breaker state detection model can be applied to actual monitoring, the sound data and vibration data of the target circuit breaker are monitored in real time, and the monitoring results are obtained through the model to determine whether the target circuit breaker is in a fault state. Through the monitoring results, the operation and maintenance personnel can timely understand the operating state of the target circuit breaker, timely discover and eliminate faults, and improve the reliability and safety of the power grid. The monitoring results can also be used to optimize the maintenance plan, extend the service life of the equipment, and reduce the maintenance cost. Through the above steps, the target circuit breaker state detection model is used to monitor the operating state of the target circuit breaker based on sound data and vibration data, and the monitoring results are obtained, which can effectively improve the operating reliability and safety of the circuit breaker.

[0035] In an embodiment of the present invention, an online monitoring method for a circuit breaker in a power grid is provided, comprising: during the operation of a target circuit breaker in the power grid, obtaining sound data and vibration data generated by the target circuit breaker, wherein the target circuit breaker is used to represent a circuit breaker that needs to be monitored online in the power grid; using a target circuit breaker state detection model to monitor the operating state of the target circuit breaker based on the sound data and the vibration data to obtain a monitoring result, wherein the target circuit breaker state detection model is used to represent a machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of the circuit breaker in a normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state. It is easy to notice that the present application can realize real-time monitoring of the operating status of the target circuit breaker by collecting the sound data and vibration data generated by the target circuit breaker and monitoring it in combination with the target circuit breaker status detection model. The monitoring method based on sound data and vibration data can more intuitively reflect the operating status of the target circuit breaker and realize online monitoring of the target circuit breaker without offline or power-off inspection of the circuit breaker. At the same time, it can also improve the accuracy and efficiency of monitoring, and can timely discover whether the target circuit breaker is in a fault state, so as to take maintenance or repair measures in time, thereby solving the technical problem of low efficiency of condition detection of circuit breakers in related technologies.

[0036] Optionally, the target circuit breaker state detection model is used to monitor the operating state of the target circuit breaker based on the sound data and the vibration data to obtain a monitoring result, including: extracting features from the sound data to obtain a sound voiceprint feature, and extracting features from the vibration data to obtain a vibration voiceprint feature; fusing the sound voiceprint feature and the vibration voiceprint feature to obtain a fused voiceprint feature; and using the target circuit breaker state detection model to monitor the operating state of the target circuit breaker based on the fused voiceprint feature to obtain a monitoring result.

[0037] In an optional embodiment, feature extraction can be performed on sound data, and sound data can be preprocessed, including filtering, noise reduction and other operations. Time domain features, frequency domain features, time-frequency features and other methods can be used to extract features from sound data to obtain sound voiceprint features; feature extraction can be performed on vibration data, and preprocessing operations can be performed on vibration data. Time domain features, frequency domain features, time-frequency features and other methods can be used to extract features from vibration data to obtain vibration voiceprint features; sound voiceprint features and vibration voiceprint features can be fused, and the obtained sound voiceprint features and vibration voiceprint features can be fused, and a feature-level fusion or decision-level fusion method can be used to obtain a fused voiceprint feature. Finally, a target circuit breaker state detection model can be used to monitor the operating state of the target circuit breaker based on the fused voiceprint features, a circuit breaker state detection model is established, and the fused voiceprint features are used as input data to monitor the operating state of the circuit breaker to obtain a monitoring result. Through the above steps, online monitoring of circuit breakers in the power grid can be achieved, and abnormal states of circuit breakers can be discovered in time.

[0038] Optionally, the method also includes: when the monitoring result shows that the target circuit breaker is not in a fault state, continuously performing online monitoring on the target circuit breaker; when the monitoring result shows that the target circuit breaker is in a fault state, obtaining the coil current signal and the moving contact travel signal of the target circuit breaker; and determining the fault type of the target circuit breaker based on the coil current signal and the moving contact travel signal.

[0039] The above-mentioned coil current signal may refer to a current signal flowing through the circuit breaker coil. The circuit breaker coil may be used to control the opening and closing action of the circuit breaker. When the coil is energized, a magnetic field is generated, thereby causing the contacts of the circuit breaker to close or separate, thereby realizing the function of opening or closing the circuit. By monitoring the current signal of the circuit breaker coil, the operating status of the circuit breaker may be understood in real time, including whether the circuit breaker is operating normally, whether there are any abnormal conditions, etc. By analyzing the waveform and size of the coil current signal, the working condition of the circuit breaker may be judged, faults may be discovered in a timely manner, and maintenance may be performed.

[0040] The above-mentioned moving contact travel signal may refer to the moving distance signal of the moving contact when the circuit breaker is opened or closed. The moving contact will move when the circuit breaker is opened or closed to open or close the circuit. By monitoring the moving contact travel signal, the switching state of the circuit breaker can be monitored in real time to determine whether the circuit breaker is working normally and whether there are any abnormal conditions in the circuit breaker. The moving contact travel signal can be obtained through sensors or other monitoring devices and transmitted to the monitoring system for real-time monitoring and analysis. By monitoring the moving contact travel signal, the state of the circuit breaker can be remotely monitored and controlled to ensure the safe operation and reliability of the power system.

[0041] In an optional embodiment, when the monitoring result shows that the target circuit breaker is not in a fault state, the target circuit breaker can be continuously monitored online. By installing sensors or monitoring equipment on the target circuit breaker, the operating status of the circuit breaker can be monitored in real time, including parameters such as coil current, number of operations, and contact stroke. The monitored data can be transmitted to a monitoring center or a cloud platform for real-time monitoring and recording through a data acquisition system. When the monitoring result shows that the target circuit breaker is in a faulty state, the coil current signal and the moving contact stroke signal of the target circuit breaker can be obtained. When the target circuit breaker is detected to have a fault, the coil current signal and the moving contact stroke signal of the circuit breaker can be collected. The coil current signal can reflect the current load of the circuit breaker, and the moving contact stroke signal can reflect the action of the contact, which is helpful to determine the cause of the fault. Then, by performing data analysis and processing on the coil current signal and the moving contact stroke signal, the fault type of the target circuit breaker can be determined. For example, when the coil current increases abnormally, it may indicate a short circuit fault, and abnormal moving contact stroke or no action may indicate a contact fault. According to different fault types, corresponding fault diagnosis and handling suggestions can be provided, as well as real-time alarm notifications to relevant personnel for handling. Through real-time monitoring and data analysis, fault diagnosis and early warning of circuit breakers can be achieved, thereby improving the safety and reliability of the power grid.

[0042] Optionally, the method also includes: obtaining sample sound data and sample vibration data of at least one circuit breaker in normal operating state; constructing an initial voiceprint feature vector of a target circuit breaker state detection model based on the sample sound data and the sample vibration data; performing dimensionality reduction processing on the initial voiceprint feature vector to obtain a target voiceprint feature vector, wherein the intra-class distance of the target voiceprint feature vector is smaller than the intra-class distance of the initial voiceprint feature vector, and the inter-class distance of the target voiceprint feature vector is larger than the inter-class distance of the initial voiceprint feature vector; and constructing a target circuit breaker state detection model based on the target voiceprint feature vector.

[0043] In an optional embodiment, sample sound data and sample vibration data of at least one circuit breaker in a normal operating state can be obtained, a circuit breaker in a normal operating state can be selected as a sample object, and the sound data and vibration data of the circuit breaker in a normal operating state can be obtained through a sound sensor and a vibration sensor; the sound data and vibration data can be preprocessed by signal processing technology to extract features, and time domain features, frequency domain features, wavelet transform and other methods can be used to extract features of the sound and vibration data to construct an initial voiceprint feature vector; the initial voiceprint feature vector can be subjected to dimensionality reduction processing to obtain a target voiceprint feature vector, and dimensionality reduction algorithms such as principal component analysis or linear discriminant analysis can be used to obtain the target voiceprint feature vector. The method reduces the dimension of the initial voiceprint feature vector to the target voiceprint feature vector, ensures that the intra-class distance of the target voiceprint feature vector is smaller than the intra-class distance of the initial voiceprint feature vector, and the inter-class distance is larger than the inter-class distance of the initial voiceprint feature vector; finally, a target circuit breaker state detection model can be constructed based on the target voiceprint feature vector, the target voiceprint feature vector can be used to construct a state detection model, and machine learning algorithms such as support vector machines and neural networks can be used for training to realize the detection and identification of the circuit breaker state. After the model training is completed, the sound and vibration signals can be monitored by real-time data, and the features can be extracted and input into the model to monitor the state of the target circuit breaker in real time and make corresponding warnings or processing.

[0044] Optionally, based on the sample sound data and the sample vibration data, constructing an initial voiceprint feature vector of the target circuit breaker state detection model, including: performing short-time Fourier transform on the sample sound data and the sample vibration data to obtain multiple time-frequency spectrum feature vectors; constructing a training sample set based on the multiple time-frequency spectrum feature vectors; based on the training sample set and a preset deep convolutional neural network, constructing an initial voiceprint feature vector, wherein the preset deep convolutional neural network is used to represent a preset basic deep convolutional neural network.

[0045] In an optional embodiment, the sample sound data and the sample vibration data can be subjected to short-time Fourier transform to obtain multiple time-frequency spectrum feature vectors, the collected sound data and vibration data can be preprocessed, including denoising, filtering and other operations, the processed data can be subjected to short-time Fourier transform to convert the time domain signal into a time-frequency domain signal to obtain multiple time-frequency spectrum feature vectors; a training sample set can be constructed based on multiple time-frequency spectrum feature vectors, and the obtained time-frequency spectrum feature vectors can be combined into a training sample set according to certain rules, and each sample can include the time-frequency spectrum feature vector of the sound data and the vibration data; an initial voiceprint feature vector can be constructed based on the training sample set and a preset deep convolutional neural network, the structure and parameters of the preset deep convolutional neural network can be designed, and then the neural network is trained using the training sample set to obtain an initial voiceprint feature vector, and during the training process, the neural network parameters can be continuously adjusted to improve the accuracy and generalization ability of the network. Through the above steps, online monitoring of circuit breakers in the power grid can be realized, and accurate monitoring and prediction of the circuit breaker state can be realized by analyzing the time-frequency spectrum characteristics of the sound data and vibration data, combined with the deep convolutional neural network technology, and the reliability and safety of the power grid equipment can be improved.

[0046] Optionally, constructing an initial voiceprint feature vector based on a training sample set and a preset deep convolutional neural network includes: determining a data scale of the training sample set, wherein the data scale includes at least one of the following: a frame length and a frame shift of multiple time-frequency spectrum feature vectors and the number of frames for constructing multiple time-frequency spectrum feature vectors of the training sample set; determining model parameters for constructing a target circuit breaker state detection model based on the data scale, wherein the model parameters include at least one of the following: the number of layers of the target circuit breaker state detection model, the size of convolution kernels of each layer and the number of convolution kernels of each layer; constructing an initial voiceprint feature vector based on the model parameters and a preset deep convolutional neural network.

[0047] In an optional embodiment, the data scale of the training sample set can be determined, and a certain amount and quality of training sample data can be collected, including sound signals and vibration signals of circuit breakers in normal states. To determine the data scale of the training sample set, frame lengths and frame shifts of multiple time-frequency spectrum feature vectors and the number of frames of multiple time-frequency spectrum feature vectors for constructing the training sample set can be selected according to actual conditions, and an appropriate data scale can be determined by analyzing the sample data; based on the determined data scale, model parameters for constructing a target circuit breaker state detection model can be determined. The model parameters include the number of layers of the target circuit breaker state detection model, the size of each layer of convolution kernels and the number of each layer of convolution kernels. The values ​​of the model parameters can be determined based on experience or experimental results in order to build an effective detection model. Based on the model parameters and the preset deep convolutional neural network, an initial voiceprint feature vector can be constructed. By building a deep learning model and inputting the time-frequency spectrum feature vector of the training sample set, the voiceprint feature vector is obtained through the training process. The model can be trained by the back propagation algorithm to optimize the network parameters so that the model can effectively detect the state of the target circuit breaker. Through the specific implementation of the above steps, an effective online monitoring system can be constructed to monitor the state of the circuit breaker in real time and improve the safety and stability of the power grid.

[0048] The technical solution proposed in this application is explained below in combination with an optional application scenario. This application proposes an online monitoring method for high-voltage circuit breakers based on voiceprint characteristics. As an important high-voltage equipment in the power system, the high-voltage circuit breaker plays a role in fault protection and control system operation mode in the power system. Its operating status is directly related to the safety and stability of the system. Mechanical failure is the main factor of circuit breaker failure. In addition to improving its reliability during the design and manufacturing process of the circuit breaker, it is more critical to improve its maintenance and inspection level during its operation. Research on mechanical fault monitoring and diagnosis methods of high-voltage circuit breakers and formulating reasonable inspection and maintenance strategies are of great significance to improving the reliability of the power system and saving manpower and material costs for inspection and maintenance.

[0049] In the related technology, mechanical failure of circuit breakers is the main factor of failure of high-voltage circuit breakers, and most of the mechanical failures occur in the operating mechanism, mainly including refusal to operate, malfunction, jamming, breakage and other failures. However, the above failures do not occur suddenly without any signs, but gradually develop from minor failures or abnormalities. Regular maintenance tests are the main maintenance means of high-voltage circuit breakers, but the regular maintenance test interval is relatively long, generally 3 years, and if a minor failure occurs in the circuit breaker during the maintenance interval and cannot be discovered in time, it is very easy to develop into a more serious failure before the next maintenance cycle; and with the development of electronic technology and computer technology, high-voltage circuit breaker maintenance methods based on condition monitoring have been widely developed. The fault diagnosis method based on analytical models uses coil current signals or moving contact travel signals for fault diagnosis. This method can identify circuit breaker faults without relying on prior knowledge of circuit breaker operation, but in actual operation, each type of circuit breaker needs to be modeled separately, and the applicability of the method is poor; based on prior The knowledge-based fault diagnosis method uses the factory characteristic parameters of the circuit breaker, expert experience or the circuit breaker reference signal as the standard, and extracts the actual signal for comparison with the standard signal for fault diagnosis. However, this method, including the fault diagnosis method based on the analytical model, uses the coil current signal or the moving contact travel curve, which is difficult to apply to online monitoring; the circuit breaker fault diagnosis method based on machine learning has received widespread attention with the rapid development of artificial intelligence in recent years, but although the traditional machine learning algorithms have achieved good results in the field of fault diagnosis, their diagnostic performance is greatly affected by the feature vector. They can only learn from the given feature vector and cannot adaptively extract and learn features of circuit breakers of various models or under different operating environments. Moreover, the detection parameters of fault diagnosis based on machine learning only use vibration parameters, and pay less attention to voiceprint parameters. This application introduces a circuit breaker online monitoring method based on voiceprint characteristics to improve the means and accuracy of online monitoring and fault diagnosis of circuit breakers.

[0050] The present application provides a method for online monitoring of high-voltage circuit breakers based on voiceprint characteristics, which can solve the problems of insufficient online monitoring means and low accuracy of high-voltage circuit breakers in operation. Specifically, during the normal daily operation of the circuit breaker, or when conducting tests and transmission operations, sound and vibration sensors can be used to collect sound and vibration data generated during the operation of the circuit breaker in a normal state, and the voiceprint characteristics of the circuit breaker operating mechanism when the circuit breaker is in a normal state can be extracted from the collected data using a deep neural network; the vibration and sound data collected when the circuit breaker is in a normal state are used as training samples, and a model in a fully normal state is obtained based on a Gaussian mixture model, and a scoring threshold is determined; after the model is established, the sound and vibration data of the circuit breaker in action can be monitored and collected in real time, the data is subjected to the above-mentioned feature extraction, the extracted feature vector is input into the Gaussian mixture model for scoring, and after comparison with the threshold, it is determined whether the transformer is in an abnormal state; if the circuit breaker is in an abnormal state, a power outage is required for maintenance, the circuit breaker can be tested and the circuit breaker current signal and the moving contact stroke curve can be collected to determine the fault type, and the fault type is marked in the model.

[0051] The present application can convert the collected sound and vibration data into a time-frequency spectrum feature vector using a short-time Fourier transform, determine the frame length, frame shift, and the number of frames that make up a training sample; according to the scale of the acquired data, the number of model layers, the size of each convolution kernel, and the number of each convolution kernel can be determined to construct a deep convolutional neural network to preliminarily obtain a voiceprint feature vector; the obtained voiceprint feature vector can be trained using a large-margin loss function to learn features with a small intra-class distance and a large inter-class distance, so that the obtained voiceprint features are more specific.

[0052] Furthermore, the voiceprint features extracted from a certain number of normal data can be used to generate a transformer state detection model under normal operating conditions based on the mixed Gaussian model; the normal data can be scored using the obtained model, and these scores are statistically analyzed to determine the abnormality detection threshold, which can be selected to be 3 times or 6 times lower than the mean in the Gaussian model.

[0053] Finally, when the circuit breaker is actuated, the sound and vibration data of the circuit breaker are collected, and the feature quantity can be extracted. The extracted feature vector is then input into the established model. If the detection score exceeds the determined threshold, the circuit breaker is abnormal. The circuit breaker current signal and moving contact travel curve can be collected to determine the fault type. The model can also be updated and trained using a small sample learning method so that the model can accurately identify the fault type.

[0054] Compared with the related art, the online monitoring method for high-voltage circuit breakers based on voiceprint characteristics provided by the present application uses an artificial intelligence algorithm to monitor whether the circuit breaker is in an abnormal state during operation. The online monitoring method for circuit breakers based on deep neural networks and statistical learning used in the present application is a method driven by massive data. A model is established based on full forward data using an unsupervised learning method. The sound and vibration state of the circuit breaker in operation can be efficiently monitored in real time without power outages, and it can be judged more quickly whether the circuit breaker is in an abnormal state. The feature extraction of the present method adopts a method based on deep neural networks, which has strong resistance to external signal interference, high algorithm intelligence, and strong migration and adaptability for different circuit breakers and different scenarios and working conditions. The present method also combines the state quantity obtained by other detection means to assist in judging the type of circuit breaker fault, and uses a small sample learning method to update and train the model, so that the model can not only judge whether the circuit breaker is in an abnormal state, but also accurately identify the fault type.

[0055] The method provided by the present application comprises the following steps: collecting sound and vibration data during the normal operation of the circuit breaker, that is, during the operation process, and extracting features from the collected sound and vibration data using a deep neural network; a neural network structure with a two-dimensional convolutional layer as the main structure can be used, and the model input is a two-dimensional spectrogram; a short-time Fourier transform method can be used to obtain a time-frequency spectrum, and a signal processing method can be used to convert a one-dimensional time series into a two-dimensional image; the ordinate of the spectrogram obtained by the short-time Fourier transform is the frequency, and the sampling rate of the sound data of the present application is relatively high, and the obtained frequency dimension is relatively high, with 4 Taking 8kHz sampling and 100ms frame length as an example, the number of sampling points per frame is 4800, the highest analysis frequency is 24kHz, and the frequency dimension is 2400 dimensions; according to the data scale obtained, the number of model layers, the size of each convolution kernel, and the number of convolution kernels in each layer can be determined to construct a deep convolutional neural network and preliminarily obtain the voiceprint feature vector; taking a frame length of 64ms, a frame shift of 32ms, 64 frames forming a training sample, and a total sample length of 2080ms (64ms+32ms*63) as an example, the deep convolution model structure shown in Table 1 can be used:

[0056] Table 1

[0057]

[0058] In the above table, t represents the channel "expansion" multiple, c represents the number of output channels, n represents the number of repetitions, and s represents the step size. The main network structure units in the above table are composed of 3x3 convolution (Conv 3x3), depthwise convolution 3x3 (Depthwise Conv3x3), inverted residual block (Inverted residual block), pointwise convolution 1x1 (Pointwise Conv 1x1), linear GD convolution 16×1 (Linear GD Conv 16×1) and linear convolution 1x1 (Linear Conv 1x1). The specific structure of the deep convolution model can also be determined according to actual needs and is not limited here. Finally, the high-dimensional spectrogram can be reduced to a 128-dimensional voiceprint feature vector.

[0059] The obtained voiceprint feature vector can be trained using the Large-Margin Loss function. After training, the same type of data in each type of data, that is, the distance within the class is small, and the different types of data, that is, the distance between classes is large, so that the obtained voiceprint feature is more specific. At the same time, the voiceprint feature after dimensionality reduction can also accurately express the original data information. The calculation formula of the Large-Margin Loss function can be expressed as:

[0060]

[0061] in,

[0062] The motivation of the large margin loss function is to add a positive integer variable m to generate a decision margin that can more strictly constrain the discriminant inequality, namely:

[0063] ||W 1 ||||x||cos(θ 1 )≥||W 1 ||||x||cos(mθ 1 )>||W 2 ||||x||cos(θ 2 );

[0064] The subscripts 1 and 2 represent type 1 and type 2, respectively. m is a positive integer. The cosine function is monotonically decreasing in the range of 0 to π, so cos(mx) is smaller than cos(x). The larger the value of m (θ 1 will be smaller) so that the model can learn features with larger inter-class distance and smaller intra-class distance; it can make the data of different classes in the extracted voiceprint feature vector have a larger distance in the feature space, and the data within the same class have a smaller distance. The final voiceprint feature has stronger specificity while being able to accurately express the original data information.

[0065] The present application can use the vibration and sound data collected during the normal operation of the circuit breaker as training samples, obtain a model based on the normal operation data based on the Gaussian mixture model training, and determine the scoring threshold; use the extracted voiceprint feature vector based on the normal operation data of the circuit breaker to establish a model of the circuit breaker in a non-fault state, that is, in an abnormal state, based on the Gaussian Mixture Model (GMM). The formula of the multi-dimensional Gaussian mixture model can be expressed as follows:

[0066]

[0067]

[0068] in is a Gaussian function, φ i is the mixing weight, K is the mixing number, and K is the hyperparameter that needs to be given. The purpose of the mixed Gaussian model is to estimate the mean and variance of each Gaussian function and its mixing weight after the mixing number is given.

[0069] The present application can use the obtained model to score the data under normal operating conditions, conduct statistical analysis on these scores, and determine the abnormal detection threshold. The threshold can be selected to be 3 times or 6 times lower than the mean in the Gaussian model. The standard deviation can be determined according to actual needs and is not limited here. The model can be put into operation, and the sound and vibration data of the circuit breaker can be collected again. The data is subjected to feature extraction, and the extracted feature vector is input into the model for scoring. After comparison with the threshold, it is determined whether the circuit breaker is in an abnormal state. If the circuit breaker is in an abnormal state, a power outage is required for maintenance. At this time, the circuit breaker needs to be tested and the circuit breaker current signal and the moving contact travel curve need to be collected to determine the fault type and mark the fault type in the model. The model can be put into operation, and the sound and vibration data of the circuit breaker can be collected again to extract feature quantities. The extracted feature vector is input into the established model for scoring. If the score exceeds the determined threshold, it is determined that the circuit breaker is abnormal. The circuit breaker opening and closing coil current signal can be collected to reflect the circuit breaker electromagnet-related faults. The moving contact stroke curve can be collected to calculate the circuit breaker opening and closing speed, maximum stroke, opening and closing time and other key technical indicators of the circuit breaker; by collecting the above two signals, the circuit breaker fault type can be comprehensively judged, such as refusal to operate, malfunction, jamming, breakage and other types of faults. After the fault type can be determined, the large boundary loss function model is updated and trained and marked using a small sample learning method, so that the model can accurately identify the fault type.

[0070] The small sample learning process can be as follows: establish a training data set with the sound and vibration data of the circuit breaker fault, randomly extract a number of categories from the training set, such as Class C faults, and only extract a small number of samples, such as k, from each category. Use this data set as the support set (Support Set) of the model, and extract q samples of Class C from the remaining data as the query set of the model; perform feature extraction and other operations on the support set, score the query set, and repeat the operation until the score is within the threshold and the score tends to a higher stable value.

[0071] Figure 2 is a flowchart of an optional method for generating a target voiceprint feature vector according to an embodiment of the present invention. Figure 2 As shown, the target voiceprint feature vector generation process includes: step S202, obtaining sample sound data and sample vibration data of at least one circuit breaker in normal operating state; step S204, performing short-time Fourier transform on the sample sound data and sample vibration data to obtain multiple time-frequency spectrum feature vectors; step S206, constructing a training sample set based on the multiple time-frequency spectrum feature vectors; constructing an initial voiceprint feature vector based on the training sample set and a preset deep convolutional neural network; step S208, performing dimensionality reduction processing on the initial voiceprint feature vector to obtain a target voiceprint feature vector.

[0072] Figure 3 is a schematic diagram of an optional online monitoring process of a circuit breaker in a power grid according to an embodiment of the present invention, such as Figure 3As shown, the online monitoring process of the circuit breaker may include a training model, detection scoring, threshold determination and discrimination process, wherein, in the training model process, the training vibration data may be framed to obtain a vibration data frame, and the vibration data frame may be subjected to feature extraction to obtain a vibration voiceprint feature; the training sound data may be framed to obtain a sound data frame, and the sound data frame may be subjected to feature extraction to obtain a sound voiceprint feature; and then, based on the vibration voiceprint feature and the sound voiceprint feature, a multi-source signal fusion voiceprint feature may be obtained, and the multi-source signal fusion voiceprint feature may be used for model training to obtain a normal state generation model. In the detection scoring process, the training vibration data may be framed to obtain a vibration data frame, and the vibration voiceprint feature may be obtained by extracting features from the vibration data frame; the training sound data may be framed to obtain a sound data frame, and the sound data frame may be subjected to feature extraction to obtain a sound voiceprint feature; and then, based on the vibration voiceprint feature and the sound voiceprint feature, a multi-source signal fusion voiceprint feature may be obtained, and then, the normal state generation model obtained in the training model process may be used to score the multi-source signal fusion voiceprint feature obtained in the detection scoring process. In the threshold determination process, the multi-source signal fusion voiceprint features can be scored based on the normal state generation model obtained in the model training process, and then statistical analysis can be performed to obtain the abnormal detection threshold. In the discrimination process, the detection result can be obtained based on the abnormal detection threshold generated in the threshold determination process and the output result of the detection scoring process.

[0073] According to another aspect of an embodiment of the present invention, an online monitoring device for a circuit breaker in a power grid is also provided. The device can execute the online monitoring method for a circuit breaker in a power grid of the above embodiment. The specific implementation method and preferred application scenario are the same as those of the above embodiment and will not be elaborated here.

[0074] Figure 4 is a schematic diagram of an online monitoring device for a circuit breaker in a power grid according to an embodiment of the present application. Figure 4 As shown, the device includes the following: an acquisition module 402 and a monitoring module 404 .

[0075] Among them, the acquisition module is used to obtain the sound data and vibration data generated by the target circuit breaker during the operation of the target circuit breaker in the power grid, wherein the target circuit breaker is used to represent the circuit breaker that needs to be monitored online in the power grid; the monitoring module is used to monitor the operating status of the target circuit breaker based on the sound data and vibration data using the target circuit breaker state detection model to obtain the monitoring result, wherein the target circuit breaker state detection model is used to represent the machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of the circuit breaker in normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state.

[0076] Among them, the monitoring module is also used to extract features from sound data to obtain sound voiceprint features, and to extract features from vibration data to obtain vibration voiceprint features; to fuse the sound voiceprint features and the vibration voiceprint features to obtain fused voiceprint features; and to monitor the operating status of the target circuit breaker based on the fused voiceprint features using the target circuit breaker status detection model to obtain monitoring results.

[0077] Among them, the monitoring module is also used to continuously perform online monitoring on the target circuit breaker when the monitoring result shows that the target circuit breaker is not in a fault state; when the monitoring result shows that the target circuit breaker is in a fault state, obtain the coil current signal and the moving contact travel signal of the target circuit breaker; and determine the fault type of the target circuit breaker based on the coil current signal and the moving contact travel signal.

[0078] Among them, the monitoring module is also used to obtain sample sound data and sample vibration data of at least one circuit breaker in normal operating state; based on the sample sound data and sample vibration data, construct an initial voiceprint feature vector of the target circuit breaker state detection model; perform dimensionality reduction processing on the initial voiceprint feature vector to obtain a target voiceprint feature vector, wherein the intra-class distance of the target voiceprint feature vector is smaller than the intra-class distance of the initial voiceprint feature vector, and the inter-class distance of the target voiceprint feature vector is larger than the inter-class distance of the initial voiceprint feature vector, and construct a target circuit breaker state detection model based on the target voiceprint feature vector.

[0079] Among them, the monitoring module is also used to perform short-time Fourier transform on sample sound data and sample vibration data to obtain multiple time-frequency spectrum feature vectors; construct a training sample set based on the multiple time-frequency spectrum feature vectors; based on the training sample set and a preset deep convolutional neural network, construct an initial voiceprint feature vector, wherein the preset deep convolutional neural network is used to represent a pre-set basic deep convolutional neural network.

[0080] The monitoring module is further used to determine the data scale of the training sample set, wherein the data scale includes at least one of the following: frame length, frame shift and number of frames of multiple time-frequency spectrum feature vectors for constructing the training sample set; based on the data scale, determine the model parameters for constructing the target circuit breaker state detection model, wherein the model parameters include at least one of the following: the number of layers of the target circuit breaker state detection model, the size of each layer of convolution kernels and the number of each layer of convolution kernels; based on the model parameters and a preset deep convolutional neural network, construct an initial voiceprint feature vector.

[0081] An embodiment of the present application further provides an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of the present invention when running.

[0082] The above-mentioned memory may refer to a device inside a computer for storing data and programs, and may include memory, hard disk, etc., wherein the memory may be used to temporarily store running programs and data, the hard disk may be used to store programs and data for a long time, and the memory may be used to enable the computer to read and write data, and execute programs; the above-mentioned processor may be responsible for executing instructions in computer programs and performing data processing, and may be responsible for controlling and executing various operations, including arithmetic operations, logical operations, data transmission, etc.

[0083] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0084] The above-mentioned computer storage medium may refer to a medium in a computer memory used to store certain discontinuous physical quantities. Computer storage media mainly include semiconductors, magnetic cores, magnetic drums, magnetic tapes, laser disks, etc.; the stored program included in the computer-readable storage medium may be a set of instructions that can be recognized and executed by a computer, running on an electronic computer, and is an information tool that meets certain needs of people.

[0085] An embodiment of the present application further provides a computer program product, including a computer program, which implements the methods in various embodiments of the present invention when executed by a processor.

[0086] The above-mentioned computer program product may refer to a software program that has been written, tested and released and can be run on a computer or other device. The computer program product may include an application, an operating system, tool software, etc., which is used to implement specific functions or solve specific problems.

[0087] An embodiment of the present application further provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0088] The above-mentioned non-volatile computer-readable storage medium may refer to a medium for storing data. The non-volatile computer-readable storage medium can keep the data from being lost when the power is off, and can be used to store long-term data, such as operating systems, applications, and user files. The non-volatile storage medium may include hard disk drives, solid-state drives, optical disks, and flash memory storage devices, etc.

[0089] The embodiments of the present application further provide a computer program, which implements the methods in the above-mentioned embodiments of the present invention when executed by a processor.

[0090] The above-mentioned computer program may refer to a collection of instructions used to tell a computer to perform a specific task or operation. A computer program may be written by a programmer using a specific programming language and may include algorithms, data structures, logic, and control flows. A computer program may be used for a variety of purposes, including application software, operating systems, and the like.

[0091] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0092] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0093] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0094] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0095] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0096] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An online monitoring method for a circuit breaker in a power grid, characterized in that: include: During the operation of a target circuit breaker in a power grid, obtaining sound data and vibration data generated by the target circuit breaker, wherein the target circuit breaker is used to represent a circuit breaker in the power grid that needs to be monitored online; The target circuit breaker state detection model is used to monitor the operating state of the target circuit breaker based on the sound data and the vibration data to obtain a monitoring result, wherein the target circuit breaker state detection model is used to represent a machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of the circuit breaker in normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state.

2. The online monitoring method for circuit breakers in a power grid according to claim 1, characterized in that: The target circuit breaker state detection model is used to monitor the operating state of the target circuit breaker based on the sound data and the vibration data to obtain a monitoring result, including: Extracting features from the sound data to obtain sound voiceprint features, and extracting features from the vibration data to obtain vibration voiceprint features; Fusion the sound voiceprint feature and the vibration voiceprint feature to obtain a fused voiceprint feature; The target circuit breaker state detection model is used to monitor the operating state of the target circuit breaker based on the fused voiceprint feature to obtain the monitoring result.

3. The online monitoring method for circuit breakers in a power grid according to claim 2, characterized in that: The method further comprises: When the monitoring result shows that the target circuit breaker is not in the fault state, continuously performing online monitoring on the target circuit breaker; When the monitoring result indicates that the target circuit breaker is in the fault state, obtaining a coil current signal and a moving contact stroke signal of the target circuit breaker; Based on the coil current signal and the moving contact stroke signal, a fault type of the target circuit breaker is determined.

4. The online monitoring method for circuit breakers in a power grid according to claim 1, characterized in that: The method further comprises: acquiring the sample sound data and the sample vibration data of at least one circuit breaker in the normal operating state; Based on the sample sound data and the sample vibration data, construct an initial voiceprint feature vector of the target circuit breaker state detection model; Performing dimensionality reduction processing on the initial voiceprint feature vector to obtain a target voiceprint feature vector, wherein the intra-class distance of the target voiceprint feature vector is smaller than the intra-class distance of the initial voiceprint feature vector, and the inter-class distance of the target voiceprint feature vector is larger than the inter-class distance of the initial voiceprint feature vector; The target circuit breaker state detection model is constructed based on the target voiceprint feature vector.

5. The online monitoring method for circuit breakers in a power grid according to claim 4, characterized in that: Based on the sample sound data and the sample vibration data, constructing an initial voiceprint feature vector of the target circuit breaker state detection model, including: Performing short-time Fourier transform on the sample sound data and the sample vibration data to obtain a plurality of time-frequency spectrum feature vectors; Constructing a training sample set based on the multiple time-frequency spectrum feature vectors; Based on the training sample set and a preset deep convolutional neural network, the initial voiceprint feature vector is constructed, wherein the preset deep convolutional neural network is used to represent a preset basic deep convolutional neural network.

6. The online monitoring method for circuit breakers in a power grid according to claim 5, characterized in that: Based on the training sample set and the preset deep convolutional neural network, constructing the initial voiceprint feature vector includes: Determining the data scale of the training sample set, wherein the data scale includes at least one of the following: frame lengths and frame shifts of the multiple time-frequency spectrum feature vectors and the number of frames of the multiple time-frequency spectrum feature vectors used to construct the training sample set; Determining model parameters for constructing the target circuit breaker state detection model based on the data scale, wherein the model parameters include at least one of the following: the number of layers of the target circuit breaker state detection model, the size of each layer of convolution kernels, and the number of each layer of convolution kernels; Based on the model parameters and the preset deep convolutional neural network, the initial voiceprint feature vector is constructed.

7. An online monitoring device for a circuit breaker in a power grid, characterized in that: include: An acquisition module, used for acquiring sound data and vibration data generated by a target circuit breaker in a power grid during operation of the target circuit breaker, wherein the target circuit breaker is used to represent a circuit breaker in the power grid that needs to be monitored online; A monitoring module is used to monitor the operating state of the target circuit breaker based on the sound data and the vibration data using a target circuit breaker state detection model to obtain a monitoring result, wherein the target circuit breaker state detection model is used to represent a machine learning model for state monitoring of the circuit breaker obtained by pre-training based on sample sound data and sample vibration data of a circuit breaker in a normal operating state, and the monitoring result is used to indicate whether the target circuit breaker is in a fault state.

8. An electronic device, characterized in that: include: A memory storing an executable program; A processor is used to run the program, wherein when the program is run, the online monitoring method for a circuit breaker in a power grid as claimed in any one of claims 1 to 6 is executed.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the storage medium is located is controlled to execute the online monitoring method for a circuit breaker in a power grid according to any one of claims 1 to 6.

10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the online monitoring method for a circuit breaker in a power grid according to any one of claims 1 to 6.

Images